Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Cloud Digital Leader practice bank has an additional 3,070 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A retail organization is designing a machine learning solution using historical tabular data. The business team has established the following requirements for the model:
Which type of machine learning model should the organization use?
A regression model is a supervised machine learning technique designed to predict a continuous numerical value based on historical input data. In tabular machine learning workflows on Vertex AI, regression analyzes mathematical relationships between independent input attributes (known as features) and a numeric target column to output an exact quantitative estimate, such as currency amounts, temperatures, or property prices.
sum_of_money_spent, number_of_items_purchased, and demographic_profile).MAE), Root Mean Square Error (RMSE), and Root Mean Squared Logarithmic Error (RMSLE).Compared to classification models that restrict outputs to discrete buckets, regression provides the exact granularity needed for financial revenue estimation. Furthermore, because the business only needs a single numerical figure per customer based on a single row of aggregated historical attributes—rather than a multi-step temporal sequence—regression is far more streamlined and data-efficient than a full time-series forecasting model.
A binary classification model is a supervised machine learning model engineered to predict a binary outcome, which means selecting one of two mutually exclusive classes. It is specifically built to answer yes or no questions—such as whether a credit card transaction is fraudulent or legitimate, or whether a customer will sign up for a monthly subscription.
This model type cannot generate a continuous numerical value like an exact dollar amount. Instead of estimating how much a customer will spend (for example, $142.50), a binary classifier can only output a categorical probability score that assigns the record to one of two discrete labels, such as Will Spend versus Will Not Spend.
true/false or 1/0).MAE or RMSE.The business requirement asked to predict whether a customer would purchase a promotional item during an upcoming sale (Yes or No), or whether an existing user would cancel their membership account next month.
While binary classification and regression both operate on tabular data where the prediction depends solely on the feature columns in the same row, they differ fundamentally in their target data types. Choosing between them always starts with inspecting the target column: categorical two-state targets require binary classification, whereas continuous numeric targets require regression.
A time-series forecasting model is a specialized predictive model that analyzes historical temporal patterns to predict a sequence of future values over time. In retail environments, forecasting is commonly used with AutoML or Looker (via the AutoARIMA algorithm) to project daily or weekly product demand over several months so warehouses can stock inventory in advance.
The scenario requires generating a single numerical output per customer based solely on the input feature columns located within that same row. Forecasting models differ from regression in two critical ways:
MAE and RMSE, forecasting models evaluate prediction bounds using observed quantile and scaled pinball loss metrics.The organization wanted to predict a sequence of daily or weekly sales quantities for specific catalog products over the next three months to optimize supply chain inventory across multiple store locations.
Although both regression and forecasting predict continuous numerical values, distinguishing between a single-row point estimate and a multi-row temporal sequence is essential in machine learning design. When each row represents an independent entity requiring one numeric prediction, regression is the appropriate architecture.
A multi-class classification model is a machine learning algorithm designed to categorize data by predicting one specific class from three or more discrete classes. Organizations use this model type for complex categorization tasks, such as segmenting retail shoppers into distinct customer personas or assigning incoming support tickets to specific department queues.
A multi-class classification model outputs a discrete categorical label rather than a continuous numerical value. While a retail team could artificially group spending into broad tiers (such as Low Spender, Medium Spender, and High Spender), the model cannot calculate an exact, continuous dollar estimate for an individual customer's upcoming monthly expenditure.
The retail organization wanted to analyze historical purchase behaviors and demographic profiles to automatically categorize shoppers into distinct marketing personas, such as Bargain Hunter, Brand Loyalist, Seasonal Shopper, or Luxury Buyer.
Students often confuse multi-class classification with regression when categories represent numerical ranges (for example, $0-$50, $51-$100, and $100+). However, predicting a pre-defined bracket is still a classification task because the output is a discrete label; predicting the exact continuous figure on an unbroken numerical scale requires regression.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated